hey evryone, im hector, im 15 built greencube because my agents kept forgetting everything between sessions. id spend time teaching them something, theyd get it, next day we were starting from scratch. drove me crazy. its a local proxy that sits between your agent and whatever llm you use. every task it extracts what worked, tracks what the agent is good and bad at per domain, and injects that back into future tasks. if the agent thumbs-down a response it stores a correction and catches that pattern next time. everything runs locally, sqlite database on your machine, no cloud, no telemetry. works with openai, ollama, openrouter, lm studio, openclaw, anything that uses the openai api format. 145 tests, 46 modules in rust. still early, looking for honest feedback on what breaks. its free and open source, mit license. try it: mac: curl -fsSL https://greencube.world/install.sh | bash win: powershell -c "irm https://greencube.world/install.ps1 | iex" then type gc in your terminal after a few tasks to see what your agent learned. this is my first real launch. any feedback welcome.

GreenCube

Stop your AI agent from repeating mistakes

Stop your AI agent from repeating mistakes